A method and system for allocating virtual resources to each participant in joint computing

Abstract

The present invention discloses a method and system for distributing benefits among participants in a joint computation. The system includes a first virtual resource allocation module, a joint computation initiation module, a second virtual resource allocation module, and a virtual resource allocation module. The method includes S101: generating first virtual resource allocation information for a joint computation participant according to a virtual resource computation rule; S102: a joint computation initiator confirming the first virtual resource allocation information and initiating a second joint computation of full sample data; S103: adjusting the first virtual resource allocation information to obtain second virtual resource allocation information; and S104: allocating virtual resources to the joint computation participants according to the second virtual resource allocation information. The virtual resources of each joint computation participant are allocated according to actual computation, thereby improving the authenticity and accuracy of the virtual resource allocation of the joint computation.

Description

Technical Field

[0001] The present invention relates to the field of joint computing, and in particular to a method and system for allocating virtual resources to various participants in joint computing. Background Art

[0002] Joint machine learning by multiple parties is currently a popular computing method in privacy computing. This computing method breaks the constraints of traditional centralized machine learning and can protect the privacy of data providers while performing distributed learning on the model.

[0003] However, this approach involves different parties, such as the computing initiator, the computing end, the data provider, the model provider, and the software platform provider. Generally speaking, the computing initiator ultimately receives the trained algorithm model as revenue and pays for it, but there is currently no good solution for how to reasonably distribute this cost. Summary of the Invention

[0004] To address the shortcomings of the existing technology, the present invention adopts the following technical solutions to allocate virtual resources of each participant in the joint computing according to the actual computing results, thereby achieving a more realistic and accurate allocation of virtual resources in the joint computing:

[0005] A method for allocating virtual resources to participants in a joint computing process, wherein the participants include a computing terminal, a software platform provider, a model provider, and a data provider, and includes the following steps:

[0006] S101, extracting first data from sample data of a data provider according to a preset ratio, performing a first joint calculation using the first data, and determining allocation of resources among a computing terminal, a software platform provider, a model provider, and a data provider based on the first joint calculation result, thereby forming first resource allocation information;

[0007] S102: The joint computing initiator confirms the first virtual resource allocation information of the joint computing participants and starts the second joint computing of the full sample data.

[0008] S103, adjusting the first virtual resource allocation information according to the adjustment scheme and the second joint calculation result to obtain second virtual resource allocation information;

[0009] S104: Allocate virtual resources to the joint computing participants according to the second virtual resource allocation information.

[0010] The benefits of each participant in the joint computation are distributed based on the actual computation results, making the joint computation benefit distribution more realistic and accurate. When starting a joint computation, a small amount of computation can quickly yield the first result. Subsequent full computations can then yield more accurate virtual resource allocation.

[0011] Furthermore, the step S103 of adjusting the first virtual resource allocation information according to the adjustment scheme and the second joint calculation result to obtain the second virtual resource allocation information includes the following steps:

[0012] S301, adjusting the first virtual resource allocation information of the computing end to the second virtual resource allocation information of the computing end based on the computing complexity, system stability, reasonable allocation of computing nodes, and communication efficiency of the computing end during the joint computing process; system stability includes whether a downtime occurs during the joint computing process;

[0013] S302, adjusting the first virtual resource allocation information of the software platform provider to the second virtual resource allocation information of the software platform provider according to the actual operation effect of the software platform provider; the actual operation effect includes: software operation efficiency and the number of software bugs;

[0014] S303, based on the comparison result of the improved model calculation efficiency and the original model calculation efficiency, adjust the first virtual resource allocation information of the model providing end to the second virtual resource allocation information of the model providing end; the original model can be a previous generation model, or a common model provided, etc.

[0015] S304: Adjust the first virtual resource allocation information on the data provider side to the second virtual resource allocation information on the data provider side based on the actual contribution of the sample data to the algorithm model. The actual contribution is determined by comparing the convergence of the algorithm model after training the algorithm model with unused data and after training the algorithm model with used data. For example, if the model converges accurately enough after training with 10,000 sample data, then the impact of adding an additional 1,000 sample data on the model will be minimal. Therefore, the contribution of the last 1,000 sample data to the model is relatively small, and the benefit of the last 1,000 sample data can be adjusted downward.

[0016] S305: Determine second resource allocation information according to the second resource configuration information of each joint computing participant.

[0017] Furthermore, the S101 includes the following steps:

[0018] S201, a computing end includes a group of computing units, determines a virtual resource weight of each computing unit, and generates first virtual resource allocation information of the computing end according to the virtual resource weight of the computing unit;

[0019] S202: Determine first virtual resource allocation information of the software platform provider based on the software configuration level of the software platform provider; the software configuration level includes configuration complexity and depth of software function application, and is divided into three levels: simple configuration, standard configuration, and professional configuration;

[0020] S203, determining first virtual resource allocation information of the model providing end according to the model characteristics of the algorithm model;

[0021] S204, determining the virtual resource weight of the sample data, and determining first virtual resource allocation information of the data provider in combination with the virtual resource weight of the sample data;

[0022] S205 , generating first virtual resource allocation information according to the first virtual resource allocation information of the computing end, the first virtual resource allocation information of the software platform providing end, the first virtual resource allocation information of the model providing end, and the first virtual resource allocation information of the data providing end.

[0023] Furthermore, in S201, according to the formula c = α × i i / t i + β × m i Determine the virtual resource weight of each computing unit, where c Indicates the i The virtual resource weight of each computing unit, i i Indicates the computational complexity, t i Indicates the actual calculation time, m i represents the space complexity, α and β Indicates adjustment parameters.

[0024] Furthermore, in S204, according to the formula π i = q i × n i Determine the virtual resource weight of the sample data, where π i Indicates the virtual resource weight of the sample data, q i represents the quality of sample training data, n iRepresents the number of sample data; the sample data includes training data, the number of training data is related to the influence of training data on model convergence, and the number of sample data is the number of data with an influence exceeding a threshold as the number of training data; the quality of the verification data is manually assessed.

[0025] A system for allocating virtual resources to each participant in a joint computing, comprising a first virtual resource allocation module, a joint computing startup module, a second virtual resource allocation module, and a virtual resource allocation module;

[0026] The first virtual resource allocation module is configured to extract first data from the sample data of the data provider according to a preset ratio (e.g., one tenth), perform a first joint calculation using the first data, determine the allocation of the computing end, the software platform provider, the model provider, and the data provider, and form first virtual resource allocation information;

[0027] The joint calculation starting module is used to obtain the first virtual resource allocation information confirmed by the joint calculation initiator and start the joint calculation of the full amount of sample data;

[0028] The second virtual resource allocation module is configured to adjust the first virtual resource allocation information according to the adjustment scheme and the joint calculation result to obtain second virtual resource allocation information;

[0029] The virtual resource allocation module allocates virtual resources to the joint computing participants according to the second virtual resource allocation information;

[0030] The joint computing participants include computing initiators, computing terminals, data providers, model providers, and software platform providers.

[0031] The calculation initiating end is used to initiate joint calculation, obtain the algorithm model obtained by the joint calculation, and provide virtual resources to all parties involved in the calculation.

[0032] Furthermore, the computing end is used to provide computing resources for joint computing;

[0033] The data provider is used to provide various data in the joint calculation;

[0034] The model providing end is used to provide the algorithm model in the joint computing;

[0035] The software platform provider is used to provide the front-end and back-end of the joint computing software;

[0036] Furthermore, the first virtual resource allocation module includes a computing end first virtual resource allocation unit, a software platform providing end first virtual resource allocation unit, a model providing end first virtual resource allocation unit and a data providing end first virtual resource allocation unit;

[0037] The computing end first virtual resource allocation unit is configured to determine a virtual resource weight of each computing unit and generate first virtual resource allocation information of the computing end according to the virtual resource weight of the computing unit;

[0038] The first virtual resource allocation unit of the software platform provider determines first virtual resource allocation information of the software platform provider according to the software configuration level of the software platform provider;

[0039] The model providing end first virtual resource allocation unit is used to determine the model providing end first virtual resource allocation information according to the model characteristics of the algorithm model;

[0040] The first virtual resource allocation unit of the data provider determines the virtual resource weight of the sample data, and determines the first virtual resource allocation information of the data provider in combination with the virtual resource weight of the sample data;

[0041] The first virtual resource allocation module generates first virtual resource allocation information based on the first virtual resource allocation information of the first virtual resource allocation unit on the computing end, the first virtual resource allocation information of the first virtual resource allocation unit on the software platform providing end, the first virtual resource allocation information of the first virtual resource allocation unit on the model providing end, and the first virtual resource allocation information of the first virtual resource allocation unit on the data providing end.

[0042] Furthermore, the second virtual resource allocation module includes a computing-end second virtual resource allocation unit, a software platform providing-end second virtual resource allocation unit, a model providing-end second virtual resource allocation unit, and a data providing-end second virtual resource allocation unit;

[0043] The second virtual resource allocation unit of the computing end is used to determine the actual computational complexity according to the joint computation result, and adjust the first virtual resource allocation information of the computing end to the second virtual resource allocation information of the computing end according to the unfinalized system, reasonable allocation of computing nodes, and communication efficiency during the joint computation process;

[0044] The second virtual resource allocation unit of the software platform provider is configured to adjust the first virtual resource allocation information of the software platform provider to the second virtual resource allocation information of the software platform provider according to the actual operation effect of the software platform provider;

[0045] The model providing end second virtual resource allocation unit is configured to adjust the model providing end first virtual resource allocation information to the model providing end second virtual resource allocation information according to a comparison result between the improved model calculation efficiency and the original model calculation efficiency;

[0046] The second virtual resource allocation unit of the data provider is configured to adjust the first virtual resource allocation information of the data provider to the second virtual resource allocation information of the data provider according to the actual contribution of the sample data to the algorithm model;

[0047] The virtual resource allocation module determines the second resource allocation information according to the second resource configuration information of each joint computing participant.

[0048] The advantages and beneficial effects of the present invention are:

[0049] The present invention provides a method and system for allocating virtual resources to each participant in a joint calculation. By adjusting the first virtual resource allocation information and obtaining the second virtual resource allocation information through the joint calculation results and the adjustment plan, the benefits of each participant in the joint calculation are distributed according to the actual calculation results, making the distribution of the benefits of the joint calculation more real and accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a flow chart of a method for distributing benefits to each participant in a joint calculation according to the present invention.

[0051] Figure 2 It is a schematic diagram of the system structure for distributing benefits among participants in a joint calculation according to the present invention. DETAILED DESCRIPTION

[0052] The following describes the specific embodiments of the present invention in detail with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to illustrate and explain the present invention and are not intended to limit the present invention.

[0053] like Figure 1As shown, a method for distributing benefits among participants in a joint computation is provided. After a computation initiator initiates a joint computation, initial virtual resource allocation information for each participant, i.e., first virtual resource allocation information, is calculated using preset computation rules. This first virtual resource allocation information is sent to the computation initiator. If the computation initiator agrees with the first virtual resource allocation information, the computation initiator notifies each participant in the joint computation to begin the joint computation. The computation initiator receives the results of the joint computation and adjusts the first virtual resource allocation information based on the results and a preset adjustment plan to obtain adjusted virtual resource allocation information, i.e., second virtual resource allocation information. This second virtual resource allocation information is sent to the benefit initiator, so that the benefit initiator distributes the benefits to the computation end, data provider, model provider, and software platform provider, respectively, based on the second virtual resource allocation information. The first virtual resource allocation information includes: first virtual resource allocation information for the computation end, first virtual resource allocation information for the data provider, first virtual resource allocation information for the model provider, and first virtual resource allocation information for the software platform provider. The second virtual resource allocation information includes: the second virtual resource allocation information of the computing end, the second virtual resource allocation information of the data providing end, the second virtual resource allocation information of the model providing end, and the second virtual resource allocation information of the software platform providing end.

[0054] Specifically, the joint computing participants include: the computing end, the software platform provider, the model provider, and the data provider, and include the following steps:

[0055] S101 extracts first data from sample data of a data provider according to a preset ratio, performs a first joint calculation using the first data, and determines allocation of resources among a computing terminal, a software platform provider, a model provider, and a data provider based on the first joint calculation result, thereby forming first resource allocation information, including the following steps:

[0056] S201, a computing end includes a group of computing units, determines a virtual resource weight of each computing unit, and generates first virtual resource allocation information of the computing end according to the virtual resource weight of the computing unit;

[0057] According to the formula c = α × i i / t i + β × m i Determine the virtual resource weight of each computing unit, where c Indicates the i The virtual resource weight of each computing unit, i i Indicates the computational complexity, t iIndicates the actual calculation time, m i represents the space complexity, α and β Indicates adjustment parameters.

[0058] S202: Determine first virtual resource allocation information of the software platform provider based on the software configuration level of the software platform provider; the software configuration level includes configuration complexity and depth of software function application, and is divided into three levels: simple configuration, standard configuration, and professional configuration;

[0059] S203, determining first virtual resource allocation information of the model provider according to model characteristics of the algorithm model; model characteristics include: versatility of the algorithm model, professionalism of the algorithm model, ability of the algorithm model to extract data information, computational efficiency of the algorithm model, and richness of output of the algorithm model;

[0060] S204, determining the virtual resource weight of the sample data, and determining first virtual resource allocation information of the data provider in combination with the virtual resource weight of the sample data;

[0061] According to the formula π i = q i × n i Determine the virtual resource weight of the sample data, where π i Indicates the virtual resource weight of the sample data, q i represents the quality of sample training data, n i Indicates the number of sample data; sample data includes training data. The number of training data is related to the influence of training data on model convergence. The number of sample data is the number of data with an influence exceeding the threshold as the number of training data. The quality of verification data is manually assessed.

[0062] S205 , generating first virtual resource allocation information according to the first virtual resource allocation information of the computing end, the first virtual resource allocation information of the software platform providing end, the first virtual resource allocation information of the model providing end, and the first virtual resource allocation information of the data providing end.

[0063] S102: The joint computing initiator confirms the first virtual resource allocation information of the joint computing participants and starts the second joint computing of the full sample data.

[0064] S103, adjusting the first virtual resource allocation information according to the adjustment scheme and the second joint calculation result to obtain second virtual resource allocation information, including the following steps:

[0065] S301, adjusting the first virtual resource allocation information of the computing end to the second virtual resource allocation information of the computing end based on the computing complexity, system stability, reasonable allocation of computing nodes, and communication efficiency of the computing end during the joint computing process; system stability includes whether a downtime occurs during the joint computing process;

[0066] S302, adjusting the first virtual resource allocation information of the software platform provider to the second virtual resource allocation information of the software platform provider according to the actual operation effect of the software platform provider; the actual operation effect includes: software operation efficiency and the number of software bugs;

[0067] S303, adjusting the first virtual resource allocation information of the model providing end to the second virtual resource allocation information of the model providing end according to the comparison result of the improved model calculation efficiency and the original model calculation efficiency;

[0068] S304: Adjust the first virtual resource allocation information on the data provider side to the second virtual resource allocation information on the data provider side based on the actual contribution of the sample data to the algorithm model. The actual contribution is determined by comparing the convergence of the algorithm model after training the algorithm model with unused data and after training the algorithm model with used data. For example, if the model converges accurately enough after training with 10,000 sample data, then the impact of adding an additional 1,000 sample data on the model will be minimal. Therefore, the contribution of the last 1,000 sample data to the model is relatively small, and the benefit of the last 1,000 sample data can be adjusted downward.

[0069] S305: Determine second resource allocation information according to the second resource configuration information of each joint computing participant.

[0070] S104: Allocate virtual resources to the joint computing participants according to the second virtual resource allocation information.

[0071] The benefits of each participant in the joint calculation are distributed according to the actual calculation results, making the distribution of the benefits of the joint calculation more real and accurate.

[0072] like Figure 2As shown, a system for allocating virtual resources to each participant in a joint computing includes a first virtual resource allocation module, a joint computing startup module, a second virtual resource allocation module, and a virtual resource allocation module; the first virtual resource allocation module includes a first virtual resource allocation unit on the computing end, a first virtual resource allocation unit on the software platform providing end, a first virtual resource allocation unit on the model providing end, a first virtual resource allocation unit on the data providing end, a sample data acquisition unit, and a sample result acquisition unit; the second virtual resource allocation module includes a second virtual resource allocation unit on the computing end, a second virtual resource allocation unit on the software platform providing end, a second virtual resource allocation unit on the model providing end, and a second virtual resource allocation unit on the data providing end;

[0073] The first virtual resource allocation module is used to extract the first data from the sample data of the data provider according to a preset ratio (such as one tenth), use the first data to perform the first joint calculation, determine the allocation of the computing end, the software platform provider, the model provider, and the data provider, and form the first virtual resource allocation information; wherein, the first virtual resource allocation unit of the computing end is used to determine the virtual resource weight of each computing unit, and generate the first virtual resource allocation information of the computing end according to the virtual resource weight of the computing unit; the first virtual resource allocation unit of the software platform provider determines the first virtual resource allocation information of the software platform provider according to the software configuration level of the software platform provider; the first virtual resource allocation unit of the model provider determines the first virtual resource allocation information of the software platform provider according to the software configuration level of the software platform provider; the first virtual resource allocation unit of the model provider determines the first virtual resource allocation information of the software platform provider according to the software configuration level of the software platform provider. The resource allocation unit is used to determine the first virtual resource allocation information of the model provider end according to the model characteristics of the algorithm model; the first virtual resource allocation unit of the data provider end determines the virtual resource weight of the sample data, and determines the first virtual resource allocation information of the data provider end in combination with the virtual resource weight of the sample data; the first virtual resource allocation module generates the first virtual resource allocation information according to the first virtual resource allocation information of the first virtual resource allocation unit of the computing end, the first virtual resource allocation information of the first virtual resource allocation unit of the software platform provider end, the first virtual resource allocation information of the first virtual resource allocation unit of the model provider end and the first virtual resource allocation information of the first virtual resource allocation unit of the data provider end.

[0074] A joint computing startup module is used to obtain the first virtual resource allocation information confirmed by the joint computing initiator and start the joint computing of the full sample data;

[0075] The second virtual resource allocation module is used to adjust the first virtual resource allocation information according to the adjustment plan and the joint calculation result to obtain the second virtual resource allocation information; wherein, the second virtual resource allocation unit on the computing end is used to determine the actual calculation complexity according to the joint calculation result, and adjust the first virtual resource allocation information on the computing end to the second virtual resource allocation information on the computing end according to the system being unfinalized, the reasonable allocation of computing nodes, and the communication efficiency during the joint calculation process; the second virtual resource allocation unit on the software platform provider end is used to adjust the first virtual resource allocation information on the software platform provider end to the second virtual resource allocation information on the software platform provider end according to the actual operation effect of the software platform provider end; the second virtual resource allocation unit on the model provider end is used to adjust the first virtual resource allocation information on the model provider end to the second virtual resource allocation information on the model provider end according to the comparison result of the improved model calculation efficiency and the original model calculation efficiency; the second virtual resource allocation unit on the data provider end is used to adjust the first virtual resource allocation information on the data provider end to the second virtual resource allocation information on the data provider end according to the actual contribution of the sample data to the algorithm model.

[0076] The virtual resource allocation module determines the second resource allocation information according to the second virtual resource allocation information of each joint computing participant, and allocates virtual resources to the joint computing participants;

[0077] The joint computing participants include the computing initiator, computing end, data provider, model provider, and software platform provider;

[0078] The computing initiator is used to initiate joint computing, obtain the algorithm model derived from the joint computing, and provide virtual resources to all parties involved in the computing;

[0079] The computing end is used to provide computing resources for joint computing;

[0080] Data provider, used to provide various data in joint computing;

[0081] The model provider is used to provide the algorithm model for joint computing;

[0082] The software platform provider is used to provide the front-end and back-end of the joint computing software.

[0083] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some or all of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for allocating virtual resources to each participant in a joint computing process, characterized by: The joint computing participants include the computing end, software platform provider, model provider, and data provider, and include the following steps: S101, extracting first data from sample data of a data provider according to a preset ratio, performing a first joint calculation using the first data, and determining allocation of the computing terminal, the software platform provider, the model provider, and the data provider based on the first joint calculation result, thereby forming first virtual resource allocation information; The steps include: S201, a computing end includes a group of computing units, determines a virtual resource weight of each computing unit, and generates first virtual resource allocation information of the computing end according to the virtual resource weight of the computing unit; S202, determining first virtual resource allocation information of the software platform provider according to the software configuration level of the software platform provider; S203, determining first virtual resource allocation information of the model providing end according to the model characteristics of the algorithm model; S204, determining the virtual resource weight of the sample data, and determining first virtual resource allocation information of the data provider in combination with the virtual resource weight of the sample data; S205, generating first virtual resource allocation information according to the first virtual resource allocation information of the computing end, the first virtual resource allocation information of the software platform providing end, the first virtual resource allocation information of the model providing end, and the first virtual resource allocation information of the data providing end; S102: The joint computing initiator confirms the first virtual resource allocation information of the joint computing participants and starts the second joint computing of the full sample data; S103, adjusting the first virtual resource allocation information according to the adjustment scheme and the second joint calculation result to obtain second virtual resource allocation information; S104: Allocate virtual resources to the joint computing participants according to the second virtual resource allocation information.

2. The method for allocating virtual resources to each participant in a joint computing according to claim 1, characterized in that: The step S103 of adjusting the first virtual resource allocation information according to the adjustment scheme and the second joint calculation result to obtain the second virtual resource allocation information includes the following steps: S301, adjusting the first virtual resource allocation information of the computing end to the second virtual resource allocation information of the computing end according to the computational complexity, system stability, reasonable allocation of computing nodes, and communication efficiency of the computing end during the joint computing process; S302, adjusting the first virtual resource allocation information of the software platform provider to the second virtual resource allocation information of the software platform provider according to the actual operation effect of the software platform provider; S303, adjusting the first virtual resource allocation information of the model providing end to the second virtual resource allocation information of the model providing end according to the comparison result of the improved model calculation efficiency and the original model calculation efficiency; S304, adjusting the first virtual resource allocation information of the data provider to the second virtual resource allocation information of the data provider according to the actual contribution of the sample data to the algorithm model; S305: Determine second resource allocation information according to the second resource configuration information of each joint computing participant.

3. The method for allocating virtual resources to each participant in a joint computing according to claim 1, characterized in that: In S201, according to the formula γ = α × θ i / t i + β × m i Determine the virtual resource weight of each computing unit, where γ Indicates the i The virtual resource weight of each computing unit, θ i Indicates the computational complexity, t i Indicates the actual calculation time, m i represents the space complexity, α and β Indicates adjustment parameters.

4. The method for allocating virtual resources to each participant in a joint computing according to claim 1, characterized in that: In S204, according to the formula π i = q i × n i Determine the virtual resource weight of the sample data, where π i Indicates the virtual resource weight of the sample data, q i represents the quality of sample training data, n i Indicates the number of sample data; the sample data includes training data, and the number of training data is related to the influence.

5. A system for allocating virtual resources to participants in a joint computing, comprising a first virtual resource allocation module, a joint computing startup module, a second virtual resource allocation module, and a virtual resource allocation module, characterized in that: The first virtual resource allocation module is configured to extract first data from the sample data of the data provider according to a preset ratio, perform a first joint calculation using the first data, determine the allocation of the computing end, the software platform provider, the model provider, and the data provider, and form first virtual resource allocation information; The first virtual resource allocation module includes a computing end first virtual resource allocation unit, a software platform providing end first virtual resource allocation unit, a model providing end first virtual resource allocation unit and a data providing end first virtual resource allocation unit; The computing end first virtual resource allocation unit is configured to determine a virtual resource weight of each computing unit and generate first virtual resource allocation information of the computing end according to the virtual resource weight of the computing unit; The first virtual resource allocation unit of the software platform provider determines first virtual resource allocation information of the software platform provider according to the software configuration level of the software platform provider; The model providing end first virtual resource allocation unit is used to determine the model providing end first virtual resource allocation information according to the model characteristics of the algorithm model; The first virtual resource allocation unit of the data provider determines the virtual resource weight of the sample data, and determines the first virtual resource allocation information of the data provider in combination with the virtual resource weight of the sample data; The first virtual resource allocation module generates first virtual resource allocation information based on the first virtual resource allocation information of the first virtual resource allocation unit on the computing end, the first virtual resource allocation information of the first virtual resource allocation unit on the software platform providing end, the first virtual resource allocation information of the first virtual resource allocation unit on the model providing end, and the first virtual resource allocation information of the first virtual resource allocation unit on the data providing end; The joint calculation starting module is used to obtain the first virtual resource allocation information confirmed by the joint calculation initiator and start the joint calculation of the full amount of sample data; The second virtual resource allocation module is configured to adjust the first virtual resource allocation information according to the adjustment scheme and the joint calculation result to obtain second virtual resource allocation information; The virtual resource allocation module allocates virtual resources to the joint computing participants according to the second virtual resource allocation information; The joint computing participants include a computing initiator, a computing end, a data provider, a model provider, and a software platform provider.

6. A system for allocating virtual resources to participants in a joint computing process according to claim 5, characterized in that: The computing initiator is used to initiate joint computing, obtain the algorithm model obtained by the joint computing, and provide virtual resources to all parties involved in the computing; The computing end is used to provide computing resources for joint computing; The data provider is used to provide various data in the joint calculation; The model providing end is used to provide the algorithm model in the joint computing; The software platform provides a terminal for providing the front-end and back-end of the joint computing software.

7. A system for allocating virtual resources to each participant in a joint computing according to claim 5, characterized in that The second virtual resource allocation module includes a computing end second virtual resource allocation unit, a software platform providing end second virtual resource allocation unit, a model providing end second virtual resource allocation unit and a data providing end second virtual resource allocation unit; The second virtual resource allocation unit of the computing end is used to determine the actual computational complexity according to the joint computation result, and adjust the first virtual resource allocation information of the computing end to the second virtual resource allocation information of the computing end according to the unfinalized system, reasonable allocation of computing nodes, and communication efficiency during the joint computation process; The second virtual resource allocation unit of the software platform provider is configured to adjust the first virtual resource allocation information of the software platform provider to the second virtual resource allocation information of the software platform provider according to the actual operation effect of the software platform provider; The model providing end second virtual resource allocation unit is configured to adjust the model providing end first virtual resource allocation information to the model providing end second virtual resource allocation information according to a comparison result between the improved model calculation efficiency and the original model calculation efficiency; The second virtual resource allocation unit of the data provider is configured to adjust the first virtual resource allocation information of the data provider to the second virtual resource allocation information of the data provider according to the actual contribution of the sample data to the algorithm model; The virtual resource allocation module determines the second resource allocation information according to the second resource configuration information of each joint computing participant.

Images